By Clive Riddle, December 10, 2013
Vipin Gopal, PhD, Vice President of Clinical Analytics for Humana gave the opening plenary presentation last week during the Seventh National Predictive Modeling Summit in Washington, DC, providing an excellent overview of the current and future state of predictive modeling in healthcare.
Vipin summarizes the state of predictive modeling from his and Humana’s perspective as follows:
- They have seen rapid evolution as a discipline over the past decade
- There is newer and better software, data sources, hardware
- There are a lot more applications
- They have a deeper understanding of their members
- There are now more efficient and effective delivery mechanisms for model output
- Predictive Modelers will make a broader and deeper impact for in the coming years
Vipin offers as advice, these guiding principles for organizations deploying predictive modeling functions:
- Establish a set of "quick wins" to drive early results and build momentum
- Show results to bolster the business case behind making further investments
- Focus on the issues that have the most direct impact on the business
- Ensure that effort is placed on key strategic issues and pressing challenges
- Address challenges with underlying data
- Clean and streamlined data is an enabler for the creation of more effective and comprehensive analytical models
Vipin advocates that leading-edge analytics should encompass these themes:
- Focus: Are we solving the right problems?
- Nimble: Rapid analytics to respond to business needs
- Cutting-edge Methods: State-of-the-art problem solving
- Tools: Leverage advancements in the analytics marketplace
- Optimize: Maximize output of analytic resources
- Integrate: Systems approach to data, analytics and action
- Real-time: Closing the feedback loop with the most recent data
What are the key components of predictive modeling in healthcare, according to Vipin? Three things that should all work towards improved outcomes, high engagement and reduced costs: (1) Integration of Data,
Action, and Analytics; (2) Infrastructure incorporating Consistent, comprehensive datasets, Cutting edge analytic tools, and Deployment to action; and (3) Talent (predictive modeling staff and outsourced vendors.)
Vipin notes that past modeling work primarily relied on claims data, while current work aggregates multiple data sources to create an integrated view of the member for consistent and rapid analytics.
So where are we headed with data sources? Vipin reviewed these Next-Gen sources:
- Text Based (EMR; Nurses’ Notes; Call Center Transcripts)
- Devices (Remote monitoring, Smart Phones)
- Online Data (Social Media Data, Web Footprint)
And where are we headed overall? Vipin sees a broad range of applications, including Clinical, Marketing,
Financial, and Fraud Detection. He sees us mining deeper data sources, driven by a need to know our consumers better, while deploying more efficient delivery mechanisms that incorporate real-time alerts and mobile devices. The “Holy Grail” in all of this? Vipin says it is predicting and influencing consumer behavior; and we need to do this in an environment in which there is a proliferation of models and we hopefully will simultaneously see efforts to have them work in unison!